2014-02-24 5:16 GMT+01:00 Kevin Kofler <kevin.kofler@chello.at>:
It is very obvious that autokarma is NOT working. It is causing way more
breakage than direct stable pushes (or manual pushes with "too little"
karma) ever caused. If direct stable pushes (or manual pushes with "too
little" karma) were so bad a problem, then how can autokarma possibly NOT be
as bad a problem?

I expect that vast majority of the people who push bodhi updates that are broken (whether caught by bodhi karma or not) would also push the same broken updates directly to stable.  So, autokarma has a chance of stopping some bad builds, but it will not add any more bad builds.  It seems to me that it can be only an improvement in quality (though we can discuss whether it is worth the effort, sure).
 
The whole concept that an update is stable because an arbitrary number of
testers gave it a +1 makes no sense whatsoever. Automatically pushing an
update with no validation whatsoever is just suicidal.

I fully agree with you testers giving +1 is not even close to proper validation, but what alternative to get proper validation do you propose as an improvement?  Dropping autokarma would replace broken validation with no validation; that's not an improvement.

Should we do away with package reviews, and instead require writing an automated test as a condition of acceptance of a package to Fedora?  That would be a fun conversation - and even if we did make such a decision, what do do about the thousands of packages that are already in the distribution and have no such tests?
    Mirek